This is a repository copy of
Modelling the Cost Effectiveness of Interventions for
Osteoporosis: Issues to Consider
.
White Rose Research Online URL for this paper:
http://eprints.whiterose.ac.uk/96199/
Version: Accepted Version
Article:
Stevenson, M.D. and Selby, P.L. (2014) Modelling the Cost Effectiveness of Interventions
for Osteoporosis: Issues to Consider. PharmacoEconomics, 32 (8). pp. 735-743. ISSN
1170-7690
https://doi.org/10.1007/s40273-014-0156-8
[email protected] https://eprints.whiterose.ac.uk/ Reuse
Unless indicated otherwise, fulltext items are protected by copyright with all rights reserved. The copyright exception in section 29 of the Copyright, Designs and Patents Act 1988 allows the making of a single copy solely for the purpose of non-commercial research or private study within the limits of fair dealing. The publisher or other rights-holder may allow further reproduction and re-use of this version - refer to the White Rose Research Online record for this item. Where records identify the publisher as the copyright holder, users can verify any specific terms of use on the publisher’s website.
Takedown
If you consider content in White Rose Research Online to be in breach of UK law, please notify us by
Modelling the Cost Effectiveness of Interventions for Osteoporosis: Issues to Consider.
Abstract
Expenditure on treating osteoporotic fractures and on preventative intervention is considerable and
is likely to rise in forthcoming years due to the association between fracture risk and age. With
funders such as the National Institute for Health and Care Excellence and the Pharmaceutical
Benefits Advisory Committee explicitly considering cost effectiveness analyses within the process of
producing guidance it is imperative that economic models are as robust as possible. This paper
details issues that need to be considered specifically related to health technology assessments of
interventions for osteoporosis and highlights limitations within the current evidence base. A likely
direction of impact on cost effectiveness of addressing the key issues has been included alongside a
tentative categorisation of the likely level of these impacts. It is likely that cost effectiveness ratios
presented in previous models which did not address the identified issues were favourable to
interventions.
Declared conflict of interests: Professor Stevenson has received grants from the National Institute
for Health Research to undertake health technology assessments within the disease area of
osteoporosis. The department in which Professor Selby works has received research support from
Amgen who manufacture drugs for the treatment of osteoporosis but he has received no personal
financial support.
Key points for decision makers
In order to accurately estimate the cost effectiveness of screening policies costs of identifying people with osteoporosis but without a recent fracture must be included
Rigid adherence to the definition of osteoporosis, or using a single threshold value for fracture risk over a defined period of time in deciding whether to treat or not is unlikely to be optimal
approaches in terms of efficient allocation of resources
Background
The internationally agreed definition of osteoporosis is a systemic skeletal disease characterised by
low bone mass and microarchitectural deterioration of bone tissue, with a consequent increase in
bone fragility and susceptibility to fracture.1 Two thresholds of bone mineral density (BMD) have
been proposed for Caucasian women based on the T-score (the number of standard deviations from
the average bone mineral density of healthy young women).2,3 The first, osteoporosis, is defined as a
T-score of 2.5 SD or less. The second, osteopenia, is defined as a T-score between 1 and 2.5 SD.
In 2000 the cost of treating osteoporotic fractures in postmenopausal females was estimated to be
£1.5 - £1.8 billion per annum in the UK.4,5 The expenditure on hip fracture in Germany in 2003 was
6
whilst fractures in the elderly were estimated to cost $14 billion in 2002 in the USA.7
The risks of hip and vertebral fractures have been shown to be distributed exponentially in relation
to age,8 using Scottish data.9 Thus, given the ageing population in developed countries the budgetary
impact of osteoporosis, whether through preventative interventions or through treatment of
fractures is likely to increase significantly. With funding organisations such as the National Institute
for Health and Care Excellence (NICE) in England and Wales and the Pharmaceutical Benefits
Advisory Committee (PBAC) in Australia explicitly considering cost effectiveness when appraising
interventions there is a requirement for robust economic evaluations.
A recent systematic review of cost effectiveness modelling analyses of identification and treatment
of those at risk of osteoporosis or osteopenia was conducted by Müller et al.10 Twenty-four papers
published since 2006 were identified and assessed using the Philips et al checklist.11 This checklist
addresses 15 dimensions of quality which are grouped into three categories: structure; data; and
consistency. Each dimension was graded as: being adequately addressed; being inadequately
addressed; or unclear or not applicable. A simple scoring system was used with 1 point awarded for
meeting demands and -1 for not meeting demands, these results show a considerable range in the
quality of reporting, (-412 to 1413). However, typically the general quality of reporting was high with a
mode of 12, a median of 10 and a mean of 8.1 (standard deviation (SD) 4.7)
The intention of this paper is to highlight areas associated with modelling the cost effectiveness of
interventions for fracture prevention that warrant further discussion. These areas have been divided
into two broad categories: structural issues and model population issues. The remit of this invited
paper was to outline, from our experience, issues to consider when modelling the cost effectiveness
of interventions for osteoporosis. As such, no formal systematic review was undertaken and this
encountered when undertaking modelling for NICE or the National Institute for Health Research.
Both authors were actively involved in informing NICE Technology Appraisals 160 (TA160)14 and 161
(TA161)15 which provided guidance on the use of alendronate, etidronate, risedronate, raloxifene
and strontium ranelate.
Structural Issues
Including the costs of patient identification
The route by which a person is identified as being a candidate for cost effective treatment with a
pharmaceutical intervention can fundamentally change the incremental cost effectiveness ratio
(ICER) of a global strategy. Patients who have sustained a fracture that required clinical treatment
are effectively self-identifying and would result in one BMD scan per patient with a fracture; in
contrast, a patient who is asymptomatic, but who met the criteria for treatment would need to be
identified amongst a larger group of people, of whom the majority may not be cost effective to
treat, resulting in a large number of BMD scans (and expense) being required. Due to the increased
costs of effectively screening people, NICE issued separate guidance for those patients with a
previous fracture, (TA161)15 and for those without (TA160).14 This division also allowed different cost
effectiveness thresholds to be used for each group as it was believed that more caution should be
exhibited in treating, and exposing to potential adverse effects, women with asymptomatic
osteoporosis than women who had sustained a fracture.
The choice of fracture sites to include
Typically, all cost effectiveness models of interventions to prevent fractures have explicitly
incorporated discrete states for: hip; wrist; and vertebral fractures. Often remaining sites have been
considered within a homogeneous although this has the disadvantage in
that it does not distinguish between the consequences of the fracture which can be widely different,
for example between other femoral fractures and rib fractures. However, if the proportions of
different costs and disutilities are used
for such fractures, as in Schousboe et al16 then this limitation is removed.
Within the modelling used to inform TA16014 and TA161,15 Stevenson et al8 grouped: pelvis and
other femoral fractures within hip fractures; tibia and fibula fractures within proximal humerus
membership decided on the similarities of the costs incurred and the effect on utility, although on
reflection the allocation of pelvis fracture may have overstated the impact of these fractures. Such
adjustments (typically a 20% incr in the work by Stevenson et al
8
) are needed so a bias against interventions for fracture prevention is not created.
Incorporation of patient history
There is evidence to suggest that the presence of a previous fracture has the effect (albeit of finite
duration) of increasing the expected risk of further fractures.17 To incorporate such patient history
whilst maintaining the impacts of all fractures sustained into a model introduces complexity. This
would require individual patient modelling approaches or a large number of health states within a
cohort model, which would require all combinations of fracture status. Assuming only four fracture
sites, but with the requirement to know whether a fracture occurred in the current year or before
(for cost and disutility purposes) would require 28 (256) fracture health states. Kanis et al report
that over one million branches would be required were the individual patient model used within this
research, to be replaced with a decision tree which predicted the event in the forthcoming year.18
The model described in Kanis et al18 formed the backbone of subsequent models constructed by the
independent assessment group to inform the NICE TA16014 and TA161.15 Individual patient models
can, however, be computationally expensive and may require efficient coding within a computer
language / package, such as R, or the use of meta-modelling techniques,19 in order to run
probabilistic sensitivity analyses (PSA). PSA is widely recommended as it produces accurate answers
in non-linear models20 and is preferred within manufacturer submissions to NICE, with the statement
The use of model structures that limit the feasibility of probabilistic sensitivity analysis should
be clearly specified and justified .21 However, individual-based models will allow easier incorporation
of factors such as the particularly high risk of fracture in the first year following a fracture which falls
gradually afterwards.22 The level of discrepancy between individual patient models and cohort
models may however not be marked, when the individual patient model was populated with data
from a Markov-based cohort model23the results were reported to be similar.24
The use of variable or fixed BMD thresholds
Although clear definitions for osteoporosis and osteopenia exist these were based on
epidemiological rather than clinical criteria.25 As such, this does not mean that recommended BMD
sub-optimal in terms of cost effectiveness to have the same treatment recommendations for people with
osteoporosis, people with osteopenia and the remaining people. An alternative method would be to
allow variable thresholds dependent on the estimated cost per QALY. Such models which assess at
what T-Score treatment should be initiated, should produce more efficient allocation of resources,
as it allows the more severe patients to have access to treatment when an entire category may not
be cost effective to treat, or alternatively, prohibit treatment for less severe patients even though
the entire category may be most-effective to treat. These approaches have been refined further by
additional sub-division of the population by age, gender and number of potential clinical risk factors
resulting in a multitude of mutually exclusive and exhaustive health states. An example of this is the
guidance associated with TA16115 where recommendations for risedronate, etidronate and
strontium ranelate use T-Score values, such as -3.0SD or -3.5SD, which are not aligned with the
definition of osteoporosis. Historically, results presented by Müller et al10 showed that more models
considered osteoporotic patients as a non-divisible category rather than allowed T-Score thresholds
to be defined based on the estimated ICERs.
However, using variable thresholds can result in patients originally identified as requiring treatment
with a relatively inexpensive intervention being denied treatment with a more costly treatment due
to the ICER being prohibitively high. This can cause concern in the clinical community when an initial
treatment has been provided, and then no substitute offered if the patient cannot tolerate the
intervention as witnessed in the comments to the NICE technology appraisals.15,Error! Bookmark not
defined. Whilst this could be addressed by using a weighted average of interventions as undertaken
in Schousboe et al26, to ensure treatment is continued this would result in inefficient allocation of
resources. This is for two reasons (i) people who could have cost effectively received an initial
treatment may no longer receive this as the application of the weighted cost of the bundle of
treatments may result in the ICER being deemed too high and (ii) people could receive interventions
for which it is known that the ICER in isolation is too high, but it is being subsidised in a weighted
bundle of interventions due to highly cost effective interventions earlier in the sequence. In both
cases, reallocation of resources would be expected to improve societal health. The tension between
Threshold defined in terms of risk of fracture over a defined period versus subgroups of patients in
whom interventions are estimated to be cost effective when using annual transition probabilities.
The benefits and disadvantages of assuming a fracture threshold over a period, for example 15%
over ten years, at which patients should receive treatment is one of contention, although it is
recognised that a single risk value may be easier to implement clinically. Risk calculators such as
FRAX®27 provide risk of hip and major fracture over a 10 year period in a relatively simple to use
webpage.
However, in TA16014 and TA16115 NICE deemed it more appropriate to provide guidance based on a
model that included the annual risks of fractures which were supplied, in confidence, by the FRAX®
developers. Briefly, the reasons against using a single risk threshold include: non-differentiation
between fracture types in terms of costs and disutility; ignoring the effects of age on cost
effectiveness in terms of requiring nursing home care and mortality rates; omission of the increased
costs of identification in those without a fracture; the possibility that the funding agency assumes
different cost effectiveness thresholds for those with and without a fracture (as NICE did); that given
the different prices and efficacy of the interventions that each intervention would have a bespoke
risk threshold if efficient allocation of resources was the goal; and that is it unclear whether all
components of fracture risk are modifiable by pharmaceutical intervention. Further research on the
possibility of using a single threshold has been reported in a NICE Decision Support Unit Paper.28
This
absolute fracture risk (including ratio of hip to major fractures) that could robustly predict a positive
recommendation in TA16014 TA However, there may be pragmatic reasons to accept under,
or over, treatment of people were this to make treatment guidance more simple to implement and
this issue is being further explored by NICE.
Assumed length of treatment
One aspect rarely considered within cost effectiveness models is the length of time a treatment is to
be provided. Many models have assumed a finite treatment length and then assumed a residual
period in which the efficacy of fracture prevention wanes. If patients are to be treated for a longer
period then these residual benefits (which are provided without cost) would be less influential and
the ICER would increase. If treatment can only be provided for a finite period then strategies that
The longer-term efficacy of alendronate was explored in the FLEX study,29 where continued
treatment for a further five years was observed to reduce vertebral fractures by 55% but not
non-vertebral fractures relative to no further treatment. A study using expected value of sample
information techniques30 concluded that an RCT evaluating the long-term efficacy of
bisphosphonates in fracture prevention appeared to be cost effective for informing decision making
in England and Wales.31
Additionally as outlined in the Adverse Events section prolonged treatment may be detrimental to
patient well-being.
Adverse Events
Currently the adverse events associated with interventions for osteoporosis, which include
osteonecrosis of the jaw or gastrointestinal problems have often not been explicitly incorporated
within models. In the modelling undertaken to inform TA16014 and TA16115 the costs and potential
adverse effects of protein pump inhibitors were included, however the NICE Appraisal Committee
were content purely to note that osteonecrosis of the jaw was a potentially side effect that would be
unfavourable to bisphosphonates.. However, as the observational evidence base grows for current
interventions or with the advent of more efficacious, but potentially more harmful interventions
these conclusions may change, as such adverse events should be reviewed prior to any new
economic analyses.
Emerging data on atypical subtrochanteric and diaphyseal femoral fractures in patients receiving
long-term bisphosphonates (approximately 1 per 1000 person years)32 indicate that this may need to
be considered in patients modelled to receive prolonged bisphosphonate treatment. The authors are
not aware of modelling work that has incorporated such fractures.
Adherence
Müller et al10 comment on the frequent omission of the effects of non-adherence within published
models of osteoporosis interventions. Within the modelling used to inform TA16014 and TA16115
adherence was modelled crudely as either fully adherent or non-adherent, where the costs of 3
thout benefit. In the results produced non-adherence did not significantly change the ICER, as the cost implications were relatively small given
people who may remain on an intervention but be less than fully adherent. This was explored by
Hiligsmann et al33 where real-world adherence in Ireland was shown to halve the benefits associated
with treatment and double the ICER when assuming a relative risk of 1.17 in the increase in fractures
associated with patients with a medical possession ratio of less than 0.8.34 More sophisticated
methods such as those by Hiligsmann et al should be used, particularly with the advent of
interventions where adherence is maintained through the delivery mechanism (such as the
once-yearly injections for zolendronic acid) where differential adherence levels could result in marked
changes to the ICER.
Model population issues
Calculation of fracture risks for the referent woman
It is typical that clinical risk factors are reported in terms of their impact on risk per Z-Score (the SD
from the average person of that age), as for example is the effect of low BMD on fracture risk.35 Care
is needed to calculate correctly the risk values for women without risk factors having taken into
account the prevalence of clinical risk factors in the group for which an average fracture rate is
reported. An example of this is provided in Stevenson et al.8 where, adjusting only for age and
previous fracture, it was estimated that the risks of hip fracture for women aged 70-74 years, with
average BMD and no fracture were 54% lower than those for the average of women aged 70-74
years. Failure to adjust average fracture rates will lead to over-estimation of fracture rates; an error
that will increase the greater number of risk factors considered and will bias analyses in favour of
interventions for fracture prevention.
Synthesising efficacy data from randomised controlled trials
Historically, interventions have been tested in randomised controlled trials (RCTs) against placebo,
which resulted in only indirect comparisons being made between competing interventions.36
However, as placebo controlled trials become less ethical, as patients should not be randomised to
treatment of proven non-effectiveness when effective (and inexpensive due to generic formulations)
treatments are available, it is likely that head-to-head RCTs will be conducted and that network
meta-analysis would be required.37 Such analyses would require explicit assumptions to be made
exhibit a class effect, or whether there are actual differences between the efficacies of these
interventions. The uncertainty in the relative effectiveness of interventions will be greater when
compared with active interventions rather than placebo unless the numbers recruited are
significantly increased as the incremental benefit will be smaller. Synthesis of the efficacy will be
made more complex if it is believed that different components of fracture risk are less modifiable
through pharmaceutical intervention as discussed in the next section.
Application of estimated efficacy levels to overall risk level when it is possible that efficacy differs
between components of fracture risk
Typically, the pivotal RCTs that have proven treatment efficacy for each intervention have recruited
elderly, females based on the presence of low BMD and/or a previous fracture. As such, it has been
shown that interventions reduce fractures in such women but not people who do not have these
characteristics. For example, there is significant circumstantial evidence that the non-vertebral
fracture reduction efficacy of bisphosphonates and denosumab are much greater in those with
femoral neck T-score <=-2.5 compared to those with femoral neck T-score > -2.5.383940 Additionally
there is little evidence that the interventions would reduce the risks conveyed by factors (that are
contained within FRAX®) such as current smoking, relatively high alcohol consumption and
rheumatoid arthritis.
Accordingly NICE requested that analyses be undertaken allow a differential (and greater) efficacy
for risks conveyed through age, sex, low BMD and prior fracture than through other patient
characteristics. In calculating the respective efficacies associated with each group care was taken to
maintain the average efficacy reported in the RCTs. Until robust data on whether interventions have
similar efficacy across all components of risk are obtained this will remain a key area of uncertainty.
Uncertain, or dated, parameter values
Whilst the costs of treating fractures are relatively known, due to the large proportion that result in
either hospitalisation or hospital treatment there is less certainty in the values assumed for other
parameters. The disutility associated with vertebral fractures often have been derived from Swedish
data 41 and which Müller et al.10 due to the lack of a utility
value prior to the fracture. The emergence of research to identify more accurately the disutilities
Whilst data reporting the rate of mortality following a hip fracture exist,42 it is likely that a
substantial component can be attributed to comorbidity;43,44 precise and recent data are not
available. The same issues relate to estimated mortality following a clinical vertebral fracture where
data from Australia45,46 and the UK47 show a significant increase in mortality, although the proportion
judged to be causally related is uncertain and has been estimated from a Swedish hospital admission
registry.48
Admission to nursing home has both a large impact on both costs and quality of life, although the
data on the incidence attributable to a hip fracture and on utility remain uncertain. Where possible,
we recommend that the impact of these uncertainties be formally reported via expected value of
partial perfect information techniques,49 which would provide funders with further evidence when
considering whether to commission research studies.
Schwarz et al50 report that they identified no RCTs that evaluated the efficacy of pharmaceutical
interventions for males using fracture as an outcome measure, however, studies were identified that
used the surrogate outcome of BMD change. Currently treatment recommendations for men have
been generated based on inferences between the BMD changes in males and those observed in
females and thus are less certain than were data on fracture reduction available.
Relative importance of the issues considered.
The issues considered have been tabulated, in order of appearance, with the authors estimating the
direction of impact of each issue on the cost effectiveness of a specific intervention compared with
no treatment and the estimated level of impact (Table 1). The categories used are broad: favourable,
unfavourable and unknown for the direction of impact; and large, moderate, slight and unknown for
the level of impact. There is little evidence regarding the level of impact and thus there is
uncertainty in this categorisation. It is seen that the incorporation of additional factors within the
model is likely to be unfavourable to interventions in terms of cost effectiveness as the number of
unfavourable impacts are greater than the favourable ones. Whilst the level of impact for: the use of
variable BMD thresholds; the use of patient specific thresholds ; and on evaluating interventions
directly cannot be estimated a priori, addressing these issues would
allow more efficient allocation of resources.
Discussion
A large limitation to this paper is that it is, in essence, expert opinion with the inherent biases that
are associated with such work. As the authors reside in England and have undertaken significant
amounts of work for NICE the paper is likely to be UK-centric although we believe that all of the
issues raised are relevant to cost effectiveness modelling of osteoporosis interventions in all regions.
A systematic review was not undertaken and references are largely limited to those that were
already known by the authors; it is probable that newer values for some parameters exist although
these would not change the intrinsic message regarding the issues that need to be considered when
undertaking health technology assessments for interventions to prevent fracture.
Modelling the cost effectiveness of interventions for reducing the risk of fractures is complex. The
asymptomatic nature of the disease means that people with osteoporosis often remain undetected
until a fracture occurs. Additionally, BMD is measured on a continuum where people with similar
T-Scores can have different cost effectiveness ratios, and subsequent funding decisions, based on
characteristics such as age and the number of associated clinical risk factors. Whilst a single 10-year
risk of fracture threshold may be appealing there are a number of reasons why this concept is
limited.
Challenges remain in terms of data collection with pressing concerns relating to: which components
of risk can be modified by pharmaceutical interventions, with currently only age, sex, low BMD and
previous fracture being clearly linked to efficacious treatment; robust data on mortality following hip
or vertebral fractures, and the rates attributable to the fracture; robust data on nursing home
admissions which are costly; the reduction in utilities associated with fracture; and the efficacy of
interventions in males.
This paper has attempted to summarise the key issues and to provide a likely direction of any impact
in terms of cost effectiveness of incorporating features as well as the estimated level of the impact.
However, it is acknowledged that the categorisati
hard evidence available. It is recommended that the features which we believe to have at least a
moderate impact should be included in any future osteoporosis model, although as complexity of
models increase it may be that individual patient models are needed. Whilst some published models
have included the majority of these features
variable BMD thresholds, use patient specific T-Score thresholds and analyse interventions directly
rather than as a package of care in order to efficiently allocate resources.
Based on our estimate of the likely favourable or unfavourable impact of including each feature it is
likely that previous estimates of cost effectiveness are likely to be favourable to interventions;
addressing all of the issues identified within this paper is likely to increase the ICER.
Different jurisdictions may have different recommendations for whether the costs of home help, lost
productivity, or of future added life years should be included in the base case analyses. Therefore
these are not discussed within this paper although it is recommended that modellers adhere to the
reference case for each jurisdiction.
Conclusion
Whilst complex modelling work has been undertaken to estimate the cost effectiveness of
interventions to reduce fracture in patients with osteoporosis or osteopenia there remain
outstanding issues. This paper has attempted to detail the issues that need to be considered when
constructing such a cost effectiveness model, highlight the likely direction and level of impact for
methods of addressing these issues and also to highlight the gaps in the evidence base that need
strengthening in order for more robust estimates of cost effectiveness to be determined. If all of the
issues identified were adequately addressed it is likely that the resultant change in ICER would be
unfavourable to interventions.
References.
1
Consensus development conference: diagnosis, prophylaxis, and treatment of osteoporosis. Am J Med 2003;94:646 50.
2
World Health Organization. Assessment of fracture risk and its application to screening for postmenopausal osteoporosis. WHO Technical Report Series No. 843.Geneva: WHO;1994.
3 Kanis JA, Melton LJ, III, Christiansen C, Johnston CC, Khaltaev N. The diagnosis of osteoporosis. J Bone Miner
Res 1994;9:1137 41. 4
Torgerson DJ, Dolan P. The cost of treating osteoporotic fractures in the United Kingdom female population the author replies. Osteoporos Int 2000;11:551 2.
5
Burge RT, Worley D, Johansen A, Bhattacharyya S, Bose U. The cost of osteoporotic fractures in the UK: projections for 2000 2020. J Med Econ 2001;4:51 2.
6
Weyler E, Gandjour A. Socioeconomic burden of hip fractures in Germany. Gesundheitswesen 2007;69(11):601 6.
7
8
Stevenson M, Lloyd Jones M, and Papaioannou D. Vitamin K to prevent fractures in older women: systematic review and economic evaluation. Health Technol Assess 2009; 13(45) 1 134
9
Singer BR, McLauchlan GJ, Robinson CM, Christie J. Epidemiology of fractures in 15,000 adults: the influence of age and gender. J Bone Joint Surg 1998;80:243 8.
10
Müller D, Pulm J, Gandjour A. Cost-effectiveness of different strategies for selecting and treating individuals at increased risk of osteoporosis or osteopenia: A systematic review. Value in Health 15 (2012) 284-298 11 Philips Z, Ginnelly L, Sculpher M, et al. Review of guidelines for good practice in decision-analytic modelling
in health technology assessment. Health Technol Assess 2004;8:iii iv, ix xi, 1 158 12
Panichkul S, Panichkul P, Sritara C, et al. Cost-effectiveness analysis of various screening methods for osteoporosis in perimenopausal Thai women. Gynecol Obstet Invest 2006;62:89 96
13
Stevenson M, Davis S, Lloyd-Jones M, Beverley C. The clinical effectiveness and cost-effectiveness of strontium ranelate for the prevention of osteoporotic fragility fractures in postmenopausal women. Health Technol Assess 2007;11:1 134.
14
http://guidance.nice.org.uk/TA160/ Accessed October 2013
15
http://guidance.nice.org.uk/TA161/ Accessed October 2013
16 Schousboe JT, Taylor BC, Fink HA, Kane RL, Cummings SR, Orwoll ES et al. Cost-effectiveness of Bone
Densitometry Followed by Treatment of Osteoporosis in Older Men. JAMA. 2007;298(6):629-637 17
Klotzbuecher CM, Ross PD, Landsman PB, Abbott TA, III, Berger M. Patients with prior fractures have an increased risk of future fractures: a summary of the literature and statistical synthesis. J Bone Miner Res 2000;15:721 39.
18
Kanis JA, Brazier J, Stevenson M, Lloyd-Jones M, Calvert NW,. Treatment of Established Osteoporosis. Health Technol Assess 2002 (6) No 29. pp 1 1 46.
19
Stevenson MD, Oakley J, Chilcott JB. Gaussian process modelling in conjunction with individual patient simulation modelling. A case study describing the calculation of cost-effectiveness ratios for the treatment of osteoporosis. Med Decis Making 24 (2004) 89-100
20 Claxton K, Sculpher M, McCabe C, Briggs A, Akehurst R, Buxton M et al. Probabilistic sensitivity analysis for
NICE technology assessment: not an optional extra. Health Economics 2005;14:339-47. 21
http://www.nice.org.uk/media/D45/1E/GuideToMethodsTechnologyAppraisal2013.pdf Accessed October
2013 22
Johnell O, Kanis JA, Odén A, Sernbo I, Redlund-Johnell I, Petterson C et al. Fracture risk following an osteoporotic fracture. Osteoporos Int 2004; 15:175-9
23
Kanis JA, Adams J, Borgström F, Cooper C, Jönsson B, Preedy D et al. The cost-effectiveness of alendronate in the management of osteoporosis Bone 2008; 42(1) 4-15
24
Stevenson M. The population of health economic models is critical. Bone (2008) 43; 214
25 Assessment of fracture risk and its application to screening for postmenopausal osteoporosis. Report of a
WHO Study Group. World Health Organ Tech Rep Ser 1994;843:1-129 26
Schousboe JT, Gourlay M, Fink HA, Taylor BC, Orwoll ES. Barrett-Connor E et al. Cost-effectiveness of bone densitometry among Caucasian women and men without a prior fracture according to age and body weight. Osteoporos Int. 2013;24(1):163-77
27
http://www.shef.ac.uk/FRAX/ Accessed October 2013
28
Stevenson M. Assessing the feasibility of transforming the recommendation in TA160, TA161 and TA204 into absolute 10-year risk of fracture. http://www.nice.org.uk/nicemedia/live/11746/65031/65031.pdf. Accessed October 2013
29
Black DM, Schwartz AV, Ensrud KE, et al. Effects of stopping or continuing alendronate after 5 years of treatment. JAMA. 2006;296(24):2927 38.
30
Ades AE, Lu G, Claxton K. Expected values of sample information calculation in medical decision making. Med Decis Making. 2004;24:207 27.
31 Stevenson MD, Oakley JE, Lloyd Jones M, Brennan A, Compston JE, McCloskey EV, Selby PL. The
cost-effectiveness of an RCT to establish whether 5 or 10 years of bisphosphonate treatment is the better duration for women with a prior fracture. Medical Decision Making 2009; 29(6) 678-689
32
Shane E, Burr D, Abrahamsen B, Adler RA, Brown TD, Cheung AM, Cosman F, et al. Atypical subtrochanteric and diaphyseal femoral fractures: Second report of a task force of the American Society for Bone and Mineral Research. J Bone Mineral Res 2013 doi: 10.1002/jbmr [Epub ahead of print]
33
34
Huybrechts KF, Ishak KJ, Caro JJ. Assessment of compliance with osteoporosis treatment and its consequences in a managed care population. Bone 2006;38:922 8.
35
Marshall D, Johnell O, Wedel H. Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. BMJ1996;312:1254 9
36
Cranney A, Guyatt G, Griffith L, Wells G, Tugwell P, Rosen C et al. Summary of Meta-Analyses of Therapies for Postmenopausal Osteoporosis 2002 Endocrine Reviews 23(4): 570 - 578
37 Ades AE, Cliffe S. Markov Chain Monte Carlo estimation of a multi-parameter decision model: consistency of
evidence and the accurate assessment of uncertainty. Med Decis Making. 2002;22:359 71. 38
Cummings SR, Black DM, Applegate WB, Barrett-Connor E, Musliner TA et al. Effect of alendronate on risk of fracture in women with low bone density but without vertebral fractures: results from the Fracture
Intervention Trial. JAMA. 1998; 280(24):2077-82. 39
McClung MR, Miller GP, Zippel H, Bensen WG, Roux C, Adami S. Effect of risedronate on the risk of hip fracture in elderly women. Hip Intervention Program Study Group. N Engl J Med. 2001;344(5):333-40. 40
McClung MR, Boonen S, Törring O, Roux C, Rizzoli R, Bone HG et al. Effect of denosumab treatment on the risk of fractures in subgroups of women with postmenopausal osteoporosis. JBMR 2012;27 (1):211 218 41 Borgström F, Zethraeus N, Johnell O, et al. Costs and quality of life associated with osteoporosis-related
fractures in Sweden. Osteoporos Int 2006;17:637 50. 42
Oden A, Dawson A, Dere W, Johnell O, Jonsson B, Kanis J. Lifetime risk of hip fractures is underestimated. Osteoporos Int 1998;8:599 603
43
Meyer HE, Tverdal A, Falch JA, Pederden JI. Factors associated with mortality after hip fracture. Osteoporos Int 2000;11:228 32.
44
Poor G, Atkinso EJ O F WM M LJ III D hip fractures in men. Clin Orthop Relat Res1995;319:260 5. 45
Center JR, Nguyen TV, Schneider D, Sambrook PN, Eisman JA. Mortality after all major types of osteoporotic fracture in men and women: an observational study. Lancet 1999;353:878 82. 46 Cauley JA, Thompson DE, Ensrud KC, Scott JC, Black D. Risk of mortality following clinical fractures.
Osteoporos Int 2000;11:556 61. 47
Jalava T, Sama S, Pylkkanen L, Mawer B, Kanis JA, Selby P, et al. Association between vertebral fracture and increased mortality in osteoporotic patients. J Bone Miner Res 2003;18:1254 60.
48
Kanis JA, Oden A, Johnell O, De Laet C, Jonsson B. Excess mortality after hospitalisation for vertebral fracture. Osteoporos Int 2004;15:108 12.
49
Felli JC,Hazen GB. Sensitivity analysis and the expected value of perfect information. Medical Decision Making. 1998; 18:95 109.
50
Schwarz P, Jorgensen NR, Mosekilde L, Vestergaard P. The evidence for efficacy of osteoporosis treatment in men with primary osteoporosis: A systematic review and meta-analysis of antiresorptive and anabolic
